AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Bias In, Bias Out? Finding Unbiased Subnetworks in Vanilla Models

Ivan Luiz De Moura Matos, Abdel Djalil Sad Saoud, Ekaterina Iakovleva, Vito Paolo Pastore, Enzo Tartaglione

CVPR 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add reweighting of the cross-entropy loss (ablation)BISE mask objective · BiasedMNIST (rho=0.99)accuracy: 91.9 -> 96.0Improved
add the mutual-information term between predicted and true bias (ablation)BISE mask objective (with reweighting) · BiasedMNIST (rho=0.99)sparsity: 18.5 -> 20.9Improved
add the mutual-information term (without reweighting) (ablation)BISE mask objective · BiasedMNIST (rho=0.99)accuracy: 91.9 -> 91.1No clear effect
add the mutual-information term (ablation)BISE mask objective · BiasedMNIST (rho=0.99)sparsity: 5.8 -> 18.7Improved
add finetuning of the extracted subnetworkBISE · BiasedMNIST (rho=0.997)accuracy: 90.8 -> 95.9Improved
add finetuning of the extracted subnetworkBISE · BiasedMNIST (rho=0.999)accuracy: 15.7 -> 60.5Improved
add finetuning of the extracted subnetworkBISE (last) · Corrupted-CIFAR10 (rho=0.995)accuracy: 21.94 -> 34.18Improved
add finetuning of the extracted subnetworkBISE · CelebAaccuracy: 89.7 -> 91.8Improved
add finetuning of the extracted subnetworkBISE · CivilCommentsworst-group accuracy: 80.4 -> 81.0Improved
extract subnetwork with BISE from vanilla modelVanilla model · BiasedMNIST (rho=0.997)accuracy: 66.1 -> 90.8Improved
extract subnetwork with BISE (last) from vanilla modelVanilla model · Corrupted-CIFAR10 (rho=0.95)accuracy: 47.18 -> 55.38Improved
use pseudo bias labels instead of true bias labelsBISE · CelebAaccuracy: 89.7 -> 87.5No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy